# Chord distance matrix
coldiss(spe.dc, byrank = FALSE, diag = TRUE)
# Hellinger distance matrix
coldiss(spe.dh, byrank = FALSE, diag = TRUE)
# Log-chord distance matrix
coldiss(spe.logchord, byrank = FALSE, diag = TRUE)
Compare the four colour plots. They all represent distance or dissimilarity
matrices built upon quantitative abundance data. Are they similar?
# Jaccard dissimilarity matrix
coldiss(spe.dj, byrank = FALSE, diag = TRUE)
Compare the Jaccard plot with the previous ones. The Jaccard plot is based on
binary data. Does this influence the result? Is the difference more important than
between the various plots based on quantitative coefficients?
Although these examples deal with species data, the Doubs transect is characterized by strong ecological gradients (e.g. oxygen, nitrate content; see Chap. 2). In
such a well-defined context, it may be interesting to assume for discussion that
species are absent for similar reasons from a given section of the stream, and
compute an association matrix based on a symmetrical coefficient for comparison
purposes. Here is an example using the simple matching coefficient S 1 (presented in
Sect. 3.3.4) .
# Simple matching dissimilarity
# (called the Sokal and Michener index in ade4)
spe.s1 <- dist.binary(spe, method = 2)
coldiss(spe.s1^2, byrank = FALSE, diag = TRUE)
Compare this symmetrical dissimilarity matrix with the Jaccard matrix. Which dissimilarities
are the most affected by taking, or not, double zeros into account?
The Code It Yourself corner #1
Write several lines of code to compute Jaccard’s “coefficient of community” (S 7 )
between sites #15 and 16 of the spe data frame. Sites 15 and 16 now have row
3.3 Q Mode: Computing Dissimilarity Matrices Among Objects
45
coldiss(spe.dc, byrank = FALSE, diag = TRUE)
# Hellinger distance matrix
coldiss(spe.dh, byrank = FALSE, diag = TRUE)
# Log-chord distance matrix
coldiss(spe.logchord, byrank = FALSE, diag = TRUE)
Compare the four colour plots. They all represent distance or dissimilarity
matrices built upon quantitative abundance data. Are they similar?
# Jaccard dissimilarity matrix
coldiss(spe.dj, byrank = FALSE, diag = TRUE)
Compare the Jaccard plot with the previous ones. The Jaccard plot is based on
binary data. Does this influence the result? Is the difference more important than
between the various plots based on quantitative coefficients?
Although these examples deal with species data, the Doubs transect is characterized by strong ecological gradients (e.g. oxygen, nitrate content; see Chap. 2). In
such a well-defined context, it may be interesting to assume for discussion that
species are absent for similar reasons from a given section of the stream, and
compute an association matrix based on a symmetrical coefficient for comparison
purposes. Here is an example using the simple matching coefficient S 1 (presented in
Sect. 3.3.4) .
# Simple matching dissimilarity
# (called the Sokal and Michener index in ade4)
spe.s1 <- dist.binary(spe, method = 2)
coldiss(spe.s1^2, byrank = FALSE, diag = TRUE)
Compare this symmetrical dissimilarity matrix with the Jaccard matrix. Which dissimilarities
are the most affected by taking, or not, double zeros into account?
The Code It Yourself corner #1
Write several lines of code to compute Jaccard’s “coefficient of community” (S 7 )
between sites #15 and 16 of the spe data frame. Sites 15 and 16 now have row
3.3 Q Mode: Computing Dissimilarity Matrices Among Objects
45